Senior AI Engineer

Senior Agentic AI Engineer

About The Role

Insurance software implementations are among the most complex, document-heavy, and process-intensive programmes in enterprise technology. A single implementation can involve thousands of configuration decisions, hundreds of requirement documents, and years of delivery time. Sapiens is rebuilding how that work gets done — using production-grade AI agents that operate across the full implementation lifecycle, from pre-sales and scoping through to configuration, testing, and go-live.

Work You'll Do

Agent architecture & orchestration

Design and implement agentic systems capable of multi-step reasoning, planning, tool use, and workflow execution against complex, document-intensive implementation processesBuild stateful workflows using LangGraph or equivalent — including branching, retries, self-correction, human-in-the-loop checkpoints, and reusable orchestration patternsEngineer for long-horizon reliability — multi-step task completion, recovery from compounding errors, planning under uncertainty, and robust tool use when individual steps failBuild the reasoning behind high-stakes implementation decisions — criteria-grounded outputs, structured review patterns, and auditable rationales that delivery consultants can act on and defend

Retrieval, grounding & context engineering

Develop end-to-end RAG pipelines: ingestion, chunking, embeddings, vector and hybrid retrieval, reranking, contextual compression, and grounding strategiesEngineer memory and context management — conversational state, persistent memory, retrieval-aware context assembly, and token-efficient context selectionApply MCP-style tool and context interfaces so agents access the right information at the right time across enterprise knowledge repositories, document sources, and structured configuration data

Reliability, evaluation & safety

Implement observability and tracing for prompts, tool calls, retrieval quality, agent traces, failures, drift, latency, and production behaviourApply guardrails, safety controls, and failure-handling to reduce hallucinations in agents whose outputs practitioners act on directly in live client settingsEvaluate agents at trajectory and task level — multi-step task success, failure-mode and regression analysis, sandboxed test environments — alongside retrieval and generation quality metrics, automated checks, and human review

Integration & production craft

Build integrations with internal and external tools, APIs, enterprise systems, databases, and model providers so agents operate reliably within real delivery workflowsDeliver production-quality Python code with strong practices in testing, CI/CD, logging, versioning, and documentation; make architecture decisions that balance quality, reliability, latency, cost, and model riskTranslate ambiguous, high-complexity implementation processes into robust system logic and reusable AI patterns; stay current with advances in agentic systems and translate research into practical engineering decisions

Required Qualifications

Demonstrated depth building and shipping production agentic AI systems — we weigh shipped systems over years in a titleStrong, hands-on experience with LangGraph or equivalent agentic orchestration frameworks, including custom orchestrationDeep proficiency in Python — clean, testable, production-ready codeExperience designing and optimising end-to-end RAG systems: indexing, retrieval, reranking, grounding, and evaluationDaily working proficiency with Claude (Anthropic API) and Claude Code — you use these tools every day, not occasionallyExperience building and deploying agents on Azure AI Foundry or an equivalent enterprise cloud AI platformPractical understanding of LLM behaviour — strengths, limitations, hallucination risks, reasoning constraints, and the evaluation methods used to measure themExperience evaluating and debugging agent behaviour at trajectory and task level, not just output qualityHands-on experience with MCP-based interoperability patterns and tool-calling agent designModern software practices: testing, CI/CD, observability, tracing, and debugging for LLM-based systems in production

Preferred Qualifications

Experience with multi-agent orchestration and agent collaboration patternsFamiliarity with vector databases — Pinecone, Weaviate, Azure AI Search, OpenSearchExperience building agents that process complex, unstructured document types — contracts, RFPs, configuration files, regulatory documentsExposure to model adaptation techniques such as LoRA or QLoRAPrior work in insurance, financial services, or enterprise SaaS implementation environmentsDemonstrated habit of staying current with AI research, benchmarks, and emerging engineering patterns